When machine learns from books: What Indian law must fix

India's first AI copyright case tests how far the law can stretch to keep up with machine learning, global precedents, and the idea of transformative use.

- BY KALINDHI BHATIA

In the age of generative artificial intelligence, the tension between technological innovation and intellectual property protection is becoming increasingly pronounced. A recent lawsuit in the United States involving Anthropic PBC, a prominent AI company founded by former OpenAI employees, has brought this issue to the forefront.

The dispute centres around the use of copyrighted literary works to train large language models (LLMs) that power Anthropic's AI chatbot, Claude. This case, while rooted in American copyright law, offers important learnings for India, where similar disputes are beginning to emerge. As India's first AI copyright litigation unfolds in the ANI vs. OpenAI case, the legal frameworks governing AI training, particularly the differences between the American fair use doctrine and India's fair dealing exception, warrant closer examination.

This legal battle began when authors Andrea Bartz, Charles Graeber, and Kirk Wallace Johnson accused Anthropic of unlawfully using their copyrighted works to train Claude. They alleged that Anthropic had not only pirated books from illegal sources such as Books3 and Library Genesis but also destructively scanned millions of lawfully purchased print books to create a massive internal digital library. This library served as a foundational dataset for training the Claude model. The authors claimed these acts constituted direct copyright infringement and undermined the emerging market for licensing books for AI training purposes.

ANTHROPIC'S DEFENCE: THE US FAIR USE CLAUSE